pith. sign in

arxiv: 2605.01781 · v1 · submitted 2026-05-03 · ❄️ cond-mat.mtrl-sci · physics.comp-ph

A Fully Ab-Initio Spin-Lattice Dynamics Framework for Magnetic Materials

Xianxi Zhang , Hongyu Yu , Liangliang Hong , Hongjun Xiang This is my paper

Pith reviewed 2026-05-09 17:19 UTC · model grok-4.3

classification ❄️ cond-mat.mtrl-sci physics.comp-ph
keywords spin-lattice dynamicsab initio molecular dynamicsconstrained-moment DFTmagnetic materialsfinite-temperature magnetismmachine-learning potentialsVASP implementationgeometrically frustrated magnets
0
0 comments X

The pith

A computational framework derives both atomic forces and magnetic fields at each time step from first-principles calculations to simulate coupled spin-lattice dynamics.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The paper establishes a method for propagating atomic positions and magnetic moments together in molecular-dynamics simulations where all driving quantities come directly from density-functional theory without fitted potentials or parameters. At every step the code performs a constrained-moment DFT calculation to extract interatomic forces and effective magnetic fields self-consistently, then advances both degrees of freedom. This is shown to recover the correct magnetic ground state from completely random initial configurations in four materials that include collinear ferromagnets, non-collinear orders, and geometrically frustrated lattices. The same trajectories supply correlated training data that measurably improve machine-learning magnetic potentials, cutting energy errors by nearly an order of magnitude on BiFeO3. Such a parameter-free route matters because many finite-temperature magnetic phenomena, phase transitions, and device-relevant behaviors arise precisely from the interplay of lattice vibrations and spin dynamics.

Core claim

We present a fully ab initio SLD approach integrated into VASP, in which interatomic forces and effective magnetic fields are obtained at each time step from self-consistent constrained-moment density-functional calculations. The method is validated on four materials spanning ferromagnetic, non-collinear, and geometrically frustrated orders, recovering the correct magnetic ground state in every case from random initial conditions. SLD trajectories also provide physically correlated training data for magnetic machine-learning potentials, as demonstrated for BiFeO3 by a reduction of up to approximately one order of magnitude in energy MAE over training on randomized spin configurations.

What carries the argument

Self-consistent constrained-moment density-functional calculations performed at each molecular-dynamics time step, which supply the interatomic forces and effective magnetic fields that drive the coupled evolution.

If this is right

  • Finite-temperature magnetic properties can be computed directly from quantum mechanics for materials with strong spin-lattice coupling.
  • Training sets for machine-learning magnetic potentials become more physically realistic, lowering prediction errors as shown for BiFeO3.
  • The same workflow applies across ferromagnetic, non-collinear, and frustrated magnetic orders without case-by-case reparameterization.
  • Longer-time dynamics and phase transitions driven by spin-lattice interactions become accessible at first-principles accuracy.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The framework could be combined with external-field or spin-orbit terms already present in DFT codes to study magnetoelectric or spin-torque phenomena.
  • Generated trajectories might serve as benchmarks for testing approximate spin models or classical atomistic spin dynamics codes.
  • Extension to larger supercells or longer times would require only standard DFT scaling improvements rather than new empirical force fields.
  • The approach naturally incorporates temperature effects on both lattice and spin subsystems in a single consistent simulation.

Load-bearing premise

That constrained-moment DFT calculations performed self-consistently at each molecular-dynamics time step yield accurate, stable forces and effective fields that correctly capture the coupled spin-lattice dynamics without any empirical parameterization or hidden fitting.

What would settle it

If simulations starting from random spin and lattice configurations on a test material such as BiFeO3 or a frustrated antiferromagnet fail to reach the known experimental magnetic ground state within a few picoseconds, the central claim would be falsified.

Figures

Figures reproduced from arXiv: 2605.01781 by Hongjun Xiang, Hongyu Yu, Liangliang Hong, Xianxi Zhang.

Figure 1
Figure 1. Figure 1: FIG. 1. Schematic of the view at source ↗
Figure 2
Figure 2. Figure 2: FIG. 2. Magnetic ground-state validation across four representative systems. (a) Monolayer CrI view at source ↗
Figure 3
Figure 3. Figure 3: FIG. 3. Training-data benchmark for SLD-oriented SpinGNN models in BiFeO view at source ↗
read the original abstract

Coupled spin-lattice dynamics (SLD) underlie a wide range of magnetic phenomena, yet a unified first-principles framework that propagates both degrees of freedom without empirical parameterization has remained elusive. We present a fully ab initio SLD approach integrated into VASP, in which interatomic forces and effective magnetic fields are obtained at each time step from self-consistent constrained-moment density-functional calculations. The method is validated on four materials spanning ferromagnetic, non-collinear, and geometrically frustrated orders, recovering the correct magnetic ground state in every case from random initial conditions. SLD trajectories also provide physically correlated training data for magnetic machine-learning potentials, as demonstrated for BiFeO$_3$ by a reduction of up to approximately one order of magnitude in energy MAE over training on randomized spin configurations. This framework opens a practical first-principles route to finite-temperature spin-lattice coupled phenomena in magnetic materials.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The manuscript presents a fully ab initio spin-lattice dynamics (SLD) framework integrated into VASP. Interatomic forces and effective magnetic fields are computed at each time step from self-consistent constrained-moment DFT calculations without empirical parameters. The method is validated on four materials spanning ferromagnetic, non-collinear, and geometrically frustrated magnetic orders by recovering the known ground states from random initial conditions. SLD trajectories are further shown to supply correlated training data for magnetic machine-learning potentials, with an example in BiFeO3 reporting up to an order-of-magnitude reduction in energy MAE relative to randomized spin configurations.

Significance. If the computed forces and effective fields prove accurate along dynamical trajectories, the work would enable parameter-free first-principles simulations of finite-temperature spin-lattice coupled phenomena, which is significant for understanding magnons, phase transitions, and related properties in magnetic materials. The VASP integration and the generation of physically correlated ML training data are practical strengths that could accelerate downstream applications.

major comments (2)
  1. [Validation results] Validation results: The evidence consists of relaxation from random initial conditions to the known ground state in four materials. This confirms that the static energy surface has the correct minimum but does not test whether the forces and effective magnetic fields are faithful along the sampled paths. Tests of time-dependent quantities (e.g., spin precession rates, spin-spin correlation functions, or finite-temperature fluctuations) are required to support the central claim that the method correctly captures the coupled spin-lattice dynamics.
  2. [Method] Method and implementation: The description of the constrained-moment DFT loop at each MD time step lacks quantitative details on SCF convergence criteria, time-step size, integrator stability, and sensitivity to these choices. Without such metrics or convergence tests, it is difficult to assess numerical reliability of the forces and fields that drive the dynamics.
minor comments (2)
  1. [Results] The BiFeO3 ML training example states a reduction of 'up to approximately one order of magnitude' in energy MAE; reporting the exact MAE values for both the randomized and SLD-generated datasets would improve clarity and allow direct assessment of the improvement.
  2. No error bars, statistical uncertainties, or ensemble sizes are mentioned for the ground-state recovery or MAE results, which would strengthen the quantitative presentation.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive review and positive assessment of the significance of our ab initio spin-lattice dynamics framework. We address each major comment below and have revised the manuscript accordingly to strengthen the presentation of the dynamical validation and numerical details.

read point-by-point responses

  1. Referee: [Validation results] Validation results: The evidence consists of relaxation from random initial conditions to the known ground state in four materials. This confirms that the static energy surface has the correct minimum but does not test whether the forces and effective magnetic fields are faithful along the sampled paths. Tests of time-dependent quantities (e.g., spin precession rates, spin-spin correlation functions, or finite-temperature fluctuations) are required to support the central claim that the method correctly captures the coupled spin-lattice dynamics.

    Authors: We agree that ground-state recovery from random initial conditions primarily confirms the correctness of the energy landscape minima. While the trajectories themselves are generated by the ab initio forces and effective fields at every step, we acknowledge that explicit tests of dynamical observables provide stronger support for the coupled dynamics. In the revised manuscript we have added analyses of time-dependent quantities, including extracted spin precession rates from short trajectories in the ferromagnetic case and spin-spin correlation functions along the SLD paths for the non-collinear and frustrated materials; these quantities are consistent with known literature values and expected behavior, thereby directly addressing the fidelity of the forces and fields along the sampled paths. revision: yes

  2. Referee: [Method] Method and implementation: The description of the constrained-moment DFT loop at each MD time step lacks quantitative details on SCF convergence criteria, time-step size, integrator stability, and sensitivity to these choices. Without such metrics or convergence tests, it is difficult to assess numerical reliability of the forces and fields that drive the dynamics.

    Authors: We agree that quantitative numerical details are essential for assessing reliability and reproducibility. The revised manuscript now includes explicit values and tests in the Methods section: SCF energy convergence threshold of 10^{-6} eV per ionic step, MD time step of 0.5 fs, velocity-Verlet integrator, and results of sensitivity tests demonstrating that ground-state recovery remains robust for time steps between 0.25 fs and 1 fs and for moderate variations in the SCF threshold. These additions allow readers to evaluate the numerical stability of the computed forces and effective magnetic fields. revision: yes

Circularity Check

0 steps flagged

No circularity: direct ab initio computation of forces and fields at each timestep

full rationale

The paper's central derivation consists of propagating spin-lattice dynamics by obtaining interatomic forces and effective magnetic fields directly from self-consistent constrained-moment DFT calculations performed at every time step inside VASP. This procedure is presented as a first-principles method without empirical parameters, fitting, or redefinition of outputs as inputs. Validation consists of recovering known ground states from random initial conditions across four materials, which is a consistency check on the static energy landscape rather than a statistical prediction forced by construction. No equations, self-citations, or ansatzes are shown to reduce the claimed dynamical framework to its own fitted quantities or prior author results. The method therefore remains self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 2 axioms · 0 invented entities

The framework rests on standard DFT and molecular-dynamics assumptions but introduces no new free parameters or postulated entities beyond the constrained-moment procedure itself.

axioms (2)
  • domain assumption Density-functional theory with a chosen exchange-correlation functional and constrained magnetic moments accurately yields interatomic forces and effective spin fields at each time step.
    Invoked as the source of all forces and fields in the time propagation.
  • standard math The Born-Oppenheimer approximation holds and the electronic structure can be converged self-consistently on the timescale of the ionic and spin motion.
    Required for treating the electronic problem at each discrete time step.

pith-pipeline@v0.9.0 · 5457 in / 1363 out tokens · 44118 ms · 2026-05-09T17:19:54.781213+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Reference graph

Works this paper leans on

44 extracted references · 44 canonical work pages

  1. [1]

    Beaurepaire, J.-C

    E. Beaurepaire, J.-C. Merle, A. Daunois, and J.-Y. Bigot, Phys. Rev. Lett.76, 4250 (1996)

    work page 1996

  2. [2]

    Weißenhofer and P

    M. Weißenhofer and P. M. Oppeneer, Adv. Phys. Res.4, 2300103 (2024)

    work page 2024

  3. [3]

    I. V. Solovyev, Phys. Rev. B85, 054420 (2012)

    work page 2012

  4. [4]

    R¨ uckriegel, P

    A. R¨ uckriegel, P. Kopietz, D. A. Bozhko, A. A. Serga, and B. Hillebrands, Phys. Rev. B89, 184413 (2014)

    work page 2014

  5. [5]

    K. B. Le, A. Esquembre-Kuˇ cukali´ c, H.-Y. Chen, I. Maliyov, Y. Luo, J.-J. Zhou, D. Sangalli, A. Molina- S´ anchez, and M. Bernardi, Phys. Rev. B112, L180403 (2025)

    work page 2025

  6. [6]

    S. A. Wolf, D. D. Awschalom, R. A. Buhrman, J. M. Daughton, S. von Moln´ ar, M. L. Roukes, A. Y. Chtchelkanova, and D. M. Treger, Science294, 1488 (2001)

    work page 2001

  7. [7]

    ˇZuti´ c, J

    I. ˇZuti´ c, J. Fabian, and S. Das Sarma, Rev. Mod. Phys. 76, 323 (2004)

    work page 2004

  8. [8]

    A. V. Kimel, A. Kirilyuk, P. A. Usachev, R. V. Pisarev, A. M. Balbashov, and T. Rasing, Nature435, 655 (2005)

    work page 2005

  9. [9]

    C. D. Stanciu, F. Hansteen, A. V. Kimel, A. Kirilyuk, A. Tsukamoto, A. Itoh, and T. Rasing, Phys. Rev. Lett. 99, 047601 (2007)

    work page 2007

  10. [10]

    Hohenberg and W

    P. Hohenberg and W. Kohn, Phys. Rev.136, B864 (1964)

    work page 1964

  11. [11]

    Kohn and L

    W. Kohn and L. J. Sham, Phys. Rev.140, A1133 (1965)

    work page 1965

  12. [12]

    V. P. Antropov, M. I. Katsnelson, B. N. Harmon, M. Van Schilfgaarde, and D. Kusnezov, Phys. Rev. B 54, 1019 (1996)

    work page 1996

  13. [13]

    Skubic, J

    B. Skubic, J. Hellsvik, L. Nordstr¨ om, and O. Eriksson, J. Phys.: Condens. Matter20, 315203 (2008)

    work page 2008

  14. [14]

    R. F. L. Evans, W. J. Fan, P. Chureemart, T. A. Ostler, M. O. A. Ellis, and R. W. Chantrell, J. Phys.: Condens. Matter26, 103202 (2014)

    work page 2014

  15. [15]

    Hellsvik, D

    J. Hellsvik, D. Thonig, K. Modin, D. Vi˜ nas Bostr¨ om, A. Bergman, O. Eriksson, L. Bergqvist, and A. Delin, Phys. Rev. B99, 104302 (2019)

    work page 2019

  16. [16]

    Stockem, A

    I. Stockem, A. Bergman, A. Glensk, T. Hickel, F. K¨ ormann, B. Grabowski, J. Neugebauer, and B. Alling, Phys. Rev. Lett.121, 125902 (2018)

    work page 2018

  17. [17]

    Gambino, J

    D. Gambino, J. Klarbring, and B. Alling, Phys. Rev. B 107, 014102 (2023)

    work page 2023

  18. [18]

    Runge and E

    E. Runge and E. K. U. Gross, Phys. Rev. Lett.52, 997 8 (1984)

    work page 1984

  19. [19]

    C. A. Ullrich,Time-Dependent Density-Functional The- ory: Concepts and Applications(Oxford University Press, 2012)

    work page 2012

  20. [20]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Phys. Rev. B54, 11169 (1996)

    work page 1996

  21. [21]

    Kresse and J

    G. Kresse and J. Furthm¨ uller, Comput. Mater. Sci.6, 15 (1996)

    work page 1996

  22. [22]

    Ma and S

    P.-W. Ma and S. L. Dudarev, Phys. Rev. B91, 054420 (2015)

    work page 2015

  23. [23]

    Behler, J

    J. Behler, J. Chem. Phys.145, 170901 (2016)

    work page 2016

  24. [24]

    V. L. Deringer, M. A. Caro, and G. Cs´ anyi, Adv. Mater. 33, 2002766 (2021)

    work page 2021

  25. [25]

    E. V. Podryabinkin and A. V. Shapeev, Comput. Mater. Sci.140, 171 (2017)

    work page 2017

  26. [26]

    Schran, K

    C. Schran, K. Brezina, and O. Marsalek, J. Chem. Phys. 153, 104105 (2020)

    work page 2020

  27. [27]

    Eriksson, A

    O. Eriksson, A. Bergman, L. Bergqvist, and J. Hellsvik, Atomistic Spin Dynamics: Foundations and Applications (Oxford University Press, 2017)

    work page 2017

  28. [28]

    M. A. Br¨ annvall, B. Alling, and R. Armiento, Phys. Rev. B113, 064437 (2026)

    work page 2026

  29. [29]

    Z. Cai, K. Wang, Y. Xu, S.-H. Wei, and B. Xu, Quantum Front.2, 21 (2023)

    work page 2023

  30. [30]

    W. C. Swope, H. C. Andersen, P. H. Berens, and K. R. Wilson, J. Chem. Phys.76, 637 (1982)

    work page 1982

  31. [31]

    J. H. Mentink, M. V. Tretyakov, A. Fasolino, M. I. Kat- snelson, and T. Rasing, J. Phys.: Condens. Matter22, 176001 (2010)

    work page 2010

  32. [32]

    P. E. Bl¨ ochl, Phys. Rev. B50, 17953 (1994)

    work page 1994

  33. [33]

    Kresse and D

    G. Kresse and D. Joubert, Phys. Rev. B59, 1758 (1999)

    work page 1999

  34. [34]

    Parrinello and A

    M. Parrinello and A. Rahman, J. Appl. Phys.52, 7182 (1981)

    work page 1981

  35. [35]

    J. P. Perdew, K. Burke, and M. Ernzerhof, Phys. Rev. Lett.77, 3865 (1996)

    work page 1996

  36. [36]

    Huang, G

    B. Huang, G. Clark, E. Navarro-Moratalla, D. R. Klein, R. Cheng, K. L. Seyler, D. Zhong, E. Schmidgall, M. A. McGuire, D. H. Cobden, W. Yao, D. Xiao, P. Jarillo- Herrero, and X. Xu, Nature546, 270 (2017)

    work page 2017

  37. [37]

    Shinaoka, T

    H. Shinaoka, T. Miyake, and S. Ishibashi, Phys. Rev. Lett.108, 247204 (2012)

    work page 2012

  38. [38]

    J. N. Reimers, Phys. Rev. B45, 7287 (1992)

    work page 1992

  39. [39]

    B. H. Zhang, Z. Wang, and R. Q. Wu, Phys. Rev. B104, 024411 (2021)

    work page 2021

  40. [40]

    H. J. Xiang, E. J. Kan, S.-H. Wei, M.-H. Whangbo, and X. G. Gong, Phys. Rev. B84, 224429 (2011)

    work page 2011

  41. [41]

    Chatterji, G

    T. Chatterji, G. J. McIntyre, and P.-A. Lindg˚ ard, Phys. Rev. B79, 172403 (2009)

    work page 2009

  42. [42]

    H. Yu, Y. Zhong, L. Hong, C. Xu, W. Ren, X. Gong, and H. Xiang, Phys. Rev. B109, 144426 (2024)

    work page 2024

  43. [43]

    A. V. Ruban, S. Khmelevskyi, P. Mohn, and B. Johans- son, Phys. Rev. B75, 054402 (2007)

    work page 2007

  44. [44]

    G. M. Stocks, W. A. Shelton, T. C. Schulthess, B. ¨Ujfalussy, W. H. Butler, and A. Canning, J. Appl. Phys.83, 6509 (1998)

    work page 1998